Improve how the model is visualised as a network graph

[suvayu]

Current issues

• visual clutter for complex models
• layout is unpredictable/updates while working on it

Resolutions/Feature request

• aggregate nodes to simplify the view, also useful when zooming in or out
  • zooming in one part of the model, shows that part and related nodes in detail, while aggregating and simplifying other "distant" nodes
  • use GIS (spatial proximity) / entity group (semantically related) information to aggregate
• let the user to create and save views; this can be
  • shared with other users of the same database
  • personal, something that's not as widely useful and is only present in the local project

Thoughts

• How to deal with shared views?
  • store in the database: save the underlying aggregations/transformations as SQL view, and re-render as needed
  • separate table specific to visualisation
• How & where to save personal views?
• Maybe start by identifying the circumstances when the user wants to save a view.

[DillonJ]

This has been discussed a little here: https://github.com/spine-tools/Spine-Toolbox/issues/180

The basic issue is that it takes some time to layout elements in graph view and it would be great to be able to save the view. However, the view also depends on the particular subset of objects/relationships that are selected and visible in the graph view, so the layout is also a function of the selection.

Also, I believe the existing automatic layout algorithm can be improved a lot. What we have is already quite advanced and it works very nicely in some cases, but in many cases I think the algorithm is missing something - for example, I see many many crossovers - I believe that the layout algorithm should seek to minimise these.